Win Back Email Examples

Cancelling twelve paid SaaS subscriptions costs about forty minutes and produces better material than any swipe file. I did it across January and February 2026, kept the addresses live for six months, and logged everything that arrived after the final invoice.

The most interesting finding came from the empty rows. Five of the twelve companies never sent a single message after the cancellation receipt, which means their churned customers are now a list that exists and does nothing.

How the twelve accounts were cancelled and monitored

Every account was a genuine paid subscription held for at least two billing cycles, cancelled through the normal in-product flow rather than by letting a card lapse. Cancellation reasons were varied on purpose: some cited price, some cited a missing feature, some skipped the exit survey entirely. Each address stayed monitored for 180 days with filters off.

The products were Loom, Miro, Webflow, Zapier, Grammarly, Calendly, Typeform, Descript, Notion, Figma, Linear and Intercom. Every message was logged with the day count since cancellation, the subject line, the sender type, the offer if any, and whether the copy referenced anything specific about how the account had been used.

Two honest limits apply. One address per product cannot reveal segmentation, so a company that sent me nothing might be sending an excellent sequence to a different churn reason. And a six month window misses anything on a twelve month cycle, which matters for annual contracts.

The capture table: what arrived, when and with what offer

Twelve products, seven sequences, and one clear split between companies that treat churn as an event and companies that treat it as an ending.

Seven emails in six months across twelve companies. That is a thin set for a stage everyone claims to care about, and it tells you how much room there is if you build the sequence properly. The mechanics of doing so sit in the win back email campaign playbook.

The cancellation confirmation is the most read email in the sequence

Every one of the twelve products sent a confirmation when I cancelled, and eleven of them treated it as a receipt. That is a strange allocation of effort, because the confirmation is opened by close to everyone who receives it and the day 60 changelog is opened by a fraction.

Four things belong in it. The exact date access ends, stated as a date rather than “at the end of your billing period”. What happens to the data, including whether it is retained, exported or deleted, and on what schedule. A one click reactivation link that does not require a new card entry. And one open question about why, asked in a sentence, not a form.

Webflow’s confirmation did three of the four and skipped the question. Descript’s did all four and was 96 words long. The other ten were payment processor defaults with a legal tone and no reactivation path at all, which means a customer who cancelled by accident, and people do, has to go and find the login page themselves.

There is a formatting decision worth making here too. Send the confirmation as plain text from the same sender identity you use for onboarding, so that the churned user sees a consistent human name across their whole history with you. Products that switch from a friendly onboarding voice to a billing robot at cancellation make the earlier warmth look like a technique, and the onboarding email templates are the place to keep both voices aligned.

Whether the sequence lives in your ESP or your product messaging tool changes what is possible here. Tools built around event streams handle the reason field and the release trigger natively, which is one of the real differences in a comparison like Customer.io against Braze, and it matters more than the pricing page suggests.

Pattern one: the changelog that names what actually shipped

Webflow at day 60 and Descript at day 52 both did the same thing. They listed what had been built since I left, in plain language, with no offer attached and one link to a changelog page.

This is the strongest pattern in the set and also the most demanding to run. It works because it answers the only question a churned customer has, which is whether the reason they left still applies. It fails when the list of changes is padded with maintenance items, because a reader who left over a real gap can tell the difference between a shipped feature and a release note.

Descript’s version named three specific things and stopped. What I would copy is the restraint and the named sender. What I would avoid is sending this to the whole churn list, because someone who cancelled over price reads a feature list as irrelevant and someone who cancelled over the missing integration reads it as an insult if that integration is not on the list.

To run it properly you need the cancellation reason stored as a field, and that means capturing it at the moment of cancellation rather than reconstructing it later. Most email tools can then hold the person in a segment until a matching release ships, which is the sort of condition described in the behavioural email trigger definition.

Pattern two: the named human who asks one question

Loom’s arrived 48 hours after cancellation. Plain text, a real name in the from field, a working reply-to, and a single question about what had gone wrong. No offer, no link, no survey.

I replied. That is the entire point. A reply opens a conversation in which the person can tell you something you cannot get from a dropdown, and roughly a third of the useful churn insight I have seen at SaaS companies came from free text answers to exactly this email.

The thing to copy is the absence of a form. The thing to avoid is automating the reply, because sending a personal question and answering with a template is worse than not asking. If nobody has time to read the answers, do not send the email.

Timing matters more here than anywhere else in the sequence. At 48 hours the decision is fresh and the person still remembers the specific moment that broke it. At three weeks they will tell you it was the price, because price is what people say when they have forgotten the real reason.

Pattern three: the export deadline that is honest rather than coercive

Zapier’s day 14 email said the account data would be available for export for a fixed window and gave the date. One link, straight to the export. No upgrade pitch anywhere in the message.

Deadlines belong in win back sequences only when they are real. If the data genuinely gets deleted on a schedule, saying so is a service. If the deadline is invented to create urgency, the customer will eventually find out, and you have converted a churned user into a detractor who tells people.

What makes Zapier’s version defensible is that it gave me something useful without asking for anything. A reactivation link sat at the bottom in small text, which is where it belongs. Copy that ordering. The email exists to help, and the offer is a footnote.

The version to avoid is the one that turns retention policy into pressure copy, usually with a countdown graphic. Two of the products in the wider capture used urgency styling in dunning emails and the effect was the same both times: the message read as a system trying to extract something, which is exactly the impression a churned customer already has.

Three patterns that fail, and why the discount is the worst of them

Miro at 30 days, Typeform at 41 and Grammarly at 74 all led with money off. None of the three referenced a product change, a usage history, or a reason. They were the same email with different logos.

A blind discount fails for a structural reason rather than a copywriting one. It tells the churned customer that the price they were paying was above what the company would accept, which retroactively makes the original subscription feel like a bad deal. If they do return at 30 percent off, you have a customer who will churn again when the discount expires, and you will have taught them that churning produces a discount.

Guilt copy is the second failure. “We miss you” attributes a feeling to a company, and the reader knows a system sent it. Worse versions imply the customer did something wrong by leaving, which is the fastest route to an unsubscribe on a list that was already fragile.

The third failure is the invisible one: sending anything at all when the product has not changed. If the reason someone left is still true, an email reminding them of the product simply confirms the decision. Silence in that situation is not laziness, it is the correct call until the release notes justify a different one.

Approach	Typical send window	What it needs to work	Risk if you get it wrong
Exit question, no offer	24 to 72 hours	A human who reads and answers replies	Template replies destroy the trust the email built
Shipped feature changelog	45 to 120 days	Churn reason stored as a field plus a release feed	Generic feature lists read as noise
Honest data deadline	7 to 30 days	A real deletion or export policy	Invented urgency turns a churned user into a detractor
Blind discount	Any	Nothing, which is the problem	Resets the price anchor and trains repeat churn
Guilt or we miss you copy	Any	Nothing	Unsubscribes and spam complaints on a fragile list
Silence with no sequence	Permanent	Nothing	A warm list of people who already paid you sits unused

The five companies that sent nothing, and what it costs them

Notion, Figma, Linear and Intercom sent no win back message at all within the window, and Intercom’s only follow up came from a sales rep working an account list rather than a lifecycle sequence. Grammarly sent one discount and then stopped.

Doing nothing is a defensible choice for exactly one situation: a product where churn is almost always terminal because the job disappeared. A team that stopped designing does not need a design tool. For most SaaS categories that is not the case, and the churned list is the warmest audience the company owns. These people already understood the product, already had a workspace, and already put a card on file.

The cost is easier to see when you put a number on it. A company with 4,000 churned accounts and an average subscription of $90 a month is sitting on a list where a 2 percent reactivation is worth over $86,000 in the following year. Two emails a quarter is not a program, but it is better than the receipt-only default.

There is one legitimate reason to stay silent that has nothing to do with strategy. If your sending domain has deliverability problems, mailing a cold churn list will make them worse, because churned users complain faster than anyone. Fix the domain reputation first, then build the sequence.

The sales-led version of this gap is different and slightly worse. Intercom’s follow up came from a rep with a calendar link, six weeks after cancellation, with no reference to the cancellation at all. Somebody had pulled a list and started dialling. A churned customer who receives a cold prospecting email from the company they just left learns that the two halves of the business do not talk, and that impression is hard to reverse later. Suppress cancelled accounts from outbound sequences for at least 90 days, then route them through lifecycle instead.

Score your own win back email before you send it

Run the draft through this before it goes near a segment. Anything below four out of six needs rewriting rather than tweaking.

Two supporting habits make the rubric easier to pass. Write the subject line last, using the specific noun from the body rather than a benefit claim, and pull from the same patterns collected in the SaaS email subject line file. And keep the win back sequence in the same tool as the rest of your lifecycle program, so suppression rules apply everywhere; splitting it across systems is how a churned user ends up receiving an onboarding drip.

What to build first, in order

Build the 48 hour exit question this week. It requires one person, a plain text template, and a calendar reminder to read the replies on Fridays. It will tell you what to build next, which is the only part of this that cannot be copied from another company.

Then capture the cancellation reason as a structured field, because everything else in the sequence depends on it. Once that exists, the changelog email becomes a query rather than a campaign, and it runs itself. The same instrumentation feeds the front of the funnel too, which is why the sequence design in the onboarding email sequence playbook and the annotated onboarding teardowns use the same event names.

Leave the discount until last, or leave it out. If the only thing you can offer a churned customer is a lower price on the product they already rejected, the problem is not the email, and the rest of the program is mapped on the SaaS email marketing hub alongside the full set of lifecycle email types.